An experiment using neural networks to approximate pure functions demo

This project is a experiment to approximate pure functions using neural networks. The project provides a small interactive scripting interface allowing end users to script single parameter functions x' = f(x) and a secondary scripting interface for tweaking the network parameters. The function f(x) output is clamped between a -1 and 1 range with the network activation set to tanh end to end.

The goal of this project is to test how well neuron can approximate the given function f(x) by observing its input and output where input x is also between a -1 to 1 range.

clone this repository and run the following. This will start a small http server on port 5000.

npm install
npm start